<![CDATA[Price competition on increasingly competitive e-commerce platforms requires businesses to implement pricing strategies that are adaptive and responsive to market dynamics. Static pricing strategies have proven to be unable to accommodate changes in demand, competitor prices, and consumer perceptions in real-time. This study aims to develop a Dynamic Pricing model based on Deep Reinforcement Learning (DRL) using the Deep Q-Network (DQN) algorithm that integrates the sentiment analysis of consumer reviews with the IndoBERT model and competitor prices obtained through web scraping. The research data was collected from the Tokopedia marketplace in the electronic product category for six months (January-June 2024), including 12,450 product reviews and 3,200 snapshots of competitors' prices from 45 sellers. The fine-tuned IndoBERT model achieved an accuracy of 91.2% and an F1-score of 0.89 in the three-class sentiment classification. The results of the experiment showed that the proposed DQN model increased total revenue by 18.7%, profit margin by 14.3%, and conversion rate by 11.2% compared to the static pricing strategy. This model also outperformed rule-based pricing by 8.1% and Q-Learning tabular by 3.3% in revenue metrics. The Ablation study confirmed that the sentiment feature contributed 6.3 percentage points to the increase in revenue. This study proves that the integration of consumer sentiment signals and competitors' prices within the framework of DRL provides a more optimal and adaptive pricing strategy in the e-commerce environment.]]>
Copyrights © 2026
